Journal of Financial Economics 68 (2003) 325–351
The value of corporate voting rights and control:
A cross-country analysis$
Tatiana Nenova
The World Bank, Washington DC 20433, USA
Received 15 March 2001; received in revised form 7 March 2002
Abstract
This paper measures the value of corporate voting rights, specifically of the control block of
votes, in a sample of 661 dual-class firms in 18 countries, in 1997. A consistent measure across
countries is proposed. The measure is adjusted for takeover probability, block-holding costs,
and dividend and liquidity differences between the share classes. The value of controlblock
votes varies widely across countries. It is close to half of firm market value in South Korea,
and close to zero in Finland. The value of control-block votes is interpreted as a lower bound
for actual private benefits of the controlling shareholder. The legal environment, law
enforcement, investor protection, takeover regulations, and power-concentrating corporate
charter provisions explain 68% of the cross-country variation in the value of control-block
votes.
r 2003 Elsevier Science B.V. All rights reserved.
JEL classification: G30, G32, G34, G38, K22
Keywords: Corporate control; Voting premium; Dual-class firms; Takeover laws; Investor protection
1. Introduction
Parties in control of a corporation are in a position to extract private benefits of
control that do not accrue to dispersed shareholders. Theoretical work on private
benefits of control is well developed. Early work such as Berle and Means (1932) and
$
I am grateful to Alexandre Aganine, Simeon Djankov, Rafael La Porta, Florencio Lopez-de-Silanes,
Mike Lubrano, Elena Ranguelova, Andrei Shleifer, and Daniel Wolfenzon, for helpful suggestions and
insightful comments.
E-mail address: [email protected] (T. Nenova).
0304-405X/03/$ - see front matter r 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0304-405X(03)00069-2
326
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Jensen and Meckling (1976) emphasizes the agency problem between managers and
shareholders.
Private benefits of control by dominant firm owners are the focus of Grossman
and Hart (1988), Bebchuk (1999) and Shleifer and Wolfenzon (2002), among others.
Examples of such benefits are influence over who is elected on the Board of Directors
or in the CEO position, the power to build business empires, and the ability to
transfer assets on nonmarket terms to related parties or consume perquisites at the
expense of the firm. The corporate finance literature has proven the existence of
private benefits using diverse approaches and data. Section 2 in Shleifer and Vishny
(1997) overviews this subject matter. Examples are studies of returns around proxy
fights (Dodd and Warner 1983) and studies of legal cases (Johnson et al., 2000).
Empirical work has then proceeded to identify a measure of the magnitude of private
benefits, in studies of block trades and vote premium studies.1
This paper focuses on the vote-value approach to private benefits. The value of a
vote is linked to the extraction of private benefits by controlling shareholders in
Zingales (1994), Rydqvist (1996) and Modigliani and Perotti (1998). The logic is that
a controlling shareholder competing for control is willing to pay to minority voteowners a positive price for their votes at the time of a control contest, up to her
expected value of control. The restrictive assumptions of this argument may render
the researcher uncomfortable with the empirical link between private benefits and
vote value. In that case, vote value can permit the identification of a lower bound of
control benefits.
Empirical investigations of vote value exist only for individual countries, in view
of the country-level divergence in security-voting and legal structures. However, a
country-specific setup foregoes country- level determinants of the value of controlblock votes, such as laws and regulations, which can have important explanatory
power. This work proposes a consistent approach to cross-country data and is
careful in assuring the comparability of the vote- value measure cross-country.
The cross-country approach to vote value makes it possible to study the effect of
the differences in legal environment and investor protection on the value received by
minority investors. The issue fits into the law and finance literature summarized by
La Porta et al. (2000b). La Porta et al. (1997), (1998) provides evidence on the link
between laws and financial development, while Claessens et al. (2000) discusses the
effect of laws on ownership structures. La Porta et al. (2000a), (2002) describe the
impact of the legal environment on dividend policies and corporate valuation,
respectively. One of the early works illustrating the link between vote value and
investor protection is Modigiliani and Perotti (1997). Using the few available vote
premium measures from existing single-country studies, the authors find a strong
negative correlation between stock market development and the investment-related
1
Examples of block trade studies are Barclay and Holderness (1989), and Dyck and Zingales (2002).
Vote value studies exist for Canada (Jog and Riding, 1986; Robinson and White, 1995; Smith and
Amoako-Adu, 1995), Israel (Levy, 1982), Italy (Zingales, 1994), South Korea (Chung and Kim, 1999),
Sweden (Rydqvist, 1996; Bergstrom and Rydqvist, 1990, 1992), Switzerland (Horner, 1988), UK
(Megginson, 1990), and US (Lease, McConnell, and Mikkelson, 1983; Zingales, 1995).
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
327
laws reflected in the vote premium. The present work explores in depth the relation
between the legal framework and vote and control value.
This paper measures the total value of all votes composing the control block
(assumed to be 50% of the voting power). This quantity will be referred to as ‘‘the
value of control-block votes’’ for short. The market value of a marginal corporate
vote is calculated implicitly from multiple- and limited-voting share prices. I further
isolate sources of corporate share value not related to the vote, e.g., dividend
payment differences and liquidity differences between the two classes of shares. The
sample includes all dual-class firms listed on the 30 largest national capital markets
in the world. Table 1 summarizes the list of countries and the number of sample firms
from each country. Dual-class firms are frequent in Brazil, Canada, Denmark,
Finland, Germany, Italy, Norway, Korea, Mexico, Sweden, and Switzerland. A
smaller proportion of listed firms have dual-class structure in Australia, Chile,
France, Hong-Kong, South Africa, the UK, and the US. Dual-class firms are also
allowed by law in India, but none is covered in DATASTREAM, the database used
in this paper. Differentiation in voting rights is forbidden by law in Belgium, China,
Japan, Singapore, and Spain. In China foreign-owned shares lose their voting power.
In Spain, different classes of shares are allowed, but the voting power needs to be
proportional to cash flows. In the Netherlands nonvoting shares must be convertible
into voting. Preferred shares in Indonesia, Malaysia, Philippines, Taiwan, and
Thailand have a fixed dividend and are redeemable. In other words, they are closer
to debt than equity.
The value of control-block votes is significant in magnitude and varies widely
across countries. Control-block votes are valued at more than a quarter of company
market capitalization in Brazil, Chile, France, Italy, Mexico, and South Korea. In
contrast, the value of control-block votes in Scandinavian and Anglo-Saxon
countries is close to zero. Germanic countries such as Switzerland and Germany
exhibit a value of control-block votes of around 10% to 15% of company market
value. Since value of control-block votes is a lower bound for actual private benefits
of the controlling shareholder, the latter are even higher. In Brazil, for example, a
50% vote-owner can hold as little as one-sixth of the total cash-flow rights, but the
expected benefit from control is at least double that amount (33.3% of the value of
the firm) on average.
Control-block votes are significantly less valuable given a stricter legal
environment. To illustrate, the average (unadjusted) value of control-block votes
is 4.5% in common law countries and 25.4% in French legal origin countries, where
investor protection is weaker. This finding is consistent with the hypothesis that
parties in control of a corporation can extract private benefits to the exclusion of
dispersed shareholders. The weaker laws lower expropriation costs thus encouraging
higher control benefits, which in turn pushes up the value of control-block votes.
Quality of investor protection, takeover rules on pricing and mandatory offers,
corporate charter provisions, and the extent of law enforcement explain 68% of the
systematic variation in the value of control-block votes. The impact of investor
protection laws and control transfer regulations are of comparable magnitude, but
proper enforcement is key. For example, the value of control-block votes is 48% of
328
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Table 1
Sample composition by country
Countries are listed by stock market capitalization in 1995 in millions of US$. The third column gives the
number of firms included in the sample by country, which comply with the following conditions: (1) there
are at least two publicly traded share classes, with unequal voting per cash-flow rights; (2) both classes are
listed on a domestic exchange and traded during the period January 1 to December 31, 1997; (3) the
limited-voting class is not convertible into the multiple-voting class; and (4) neither class receives a fixed
dividend, independent of that of the other class nor is redeemable or callable at the option of the firm at a
pre-set price.
Country
United States
Japan
United Kingdom
Germany
France
Switzerland
Canada
Netherlands
Hong Kong
South Africa
Australia
Malaysia
Italy
Spain
Taiwan
Korea (Rep. Of)
Sweden
Singapore
Brazil
Thailand
India
Belgium
Mexico
Chile
Indonesia
Philippines
Denmark
Norway
Finland
China
Saudi Arabia
Argentina
Israel
Austria
New Zealand
Luxembourg
Ireland
Turkey
Stock market capitalization
6,857,622
3,667,292
1,407,737
577,365
522,053
433,621
366,344
356,481
303,705
280,526
245,218
222,729
209,522
197,788
187,206
181,955
178,049
148,004
147,636
141,507
127,515
104,960
90,694
73,860
66,585
58,859
56,233
44,587
44,138
42,055
40,961
37,783
36,399
32,513
31,950
30,443
25,817
20,772
DATASTREAM dual-class firms
39
0
27
65
9
37
65
0
5
22
3
0
62
0
0
65
43
0
141
0
0
0
5
7
0
0
30
15
21
0
0
0
0
13
2
1
0
1
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
329
firm value on average in the case of a widely held firm that operates in a ‘‘worst case’’
legal environment (i.e., where the four legal variables take their lowest in-sample
values). An improvement of law enforcement from the lowest to the highest insample
value decreases the value of control-block votes to 31%. This figure drops further
down to 20% if investor protection is raised from lowest to highest in-sample value.
The same improvement in takeover rules takes the value of control-block votes down
to 8%. Finally, moving from full to no use of power-concentrating charter
provisions takes the value of controlblock votes down to 5%.
The paper transfers focus from the value of a single vote, the voting premium, to
the value of control-block votes in the aggregate. The link between the two
magnitudes can depend on regulations in the market for corporate control, as well as
on the general legal environment and firm-specific charter provisions. I make those
factors explicit by proposing a list of institutional, regulatory, as well as firm- level
determinants of the value of control-block votes. The paper provides a
comprehensive description of several relevant regulations of the market for
corporate control, as well as a taxonomy of relevant firm- level corporate charter
provisions. The study considerably expands the number of countries where vote
premiums are measured. Finally, the results demonstrate that differences in dividend
payouts, liquidity, and other nonvote-related characteristics of the two classes of
shares have a significant effect on the measurement of vote value and failure to use
proper controls can lead to biases.
The paper proceeds as follows. Section 2 discusses the estimation of the value of
controlblock votes and identifies the appropriate control variables. Section 3
presents and tests determinants of the value of control-block votes. Section 4
concludes.
2. Valuation of control-block votes and identification of control variables
The estimation of vote value is based on the general theoretical framework put
forth by Grossman and Hart (1988) and extended for the case of dual-class firms by
several studies, such as Zingales (1994, 1995), Rydqvist (1996), and Nenova (2001).
A marginal vote does not confer any control rights to a dispersed shareholder, due
to coordination issues, asymmetric information, etc. Yet a marginal voting share will
be valued higher than the verifiable cash-flow rights it confers, provided two things.
First, that control is valuable to a large shareholder. Second, there is a positive
probability that such a block-holder will demand marginal votes in order to attain
control. In other words, vote value can be realized when the vote is sold to a
shareholder for whom control carries a positive value. Control typically means the
ownership of a sufficient voting power to take decisions on important company
matters. Here the threshold is assumed (without loss of generality) to be 50% of the
votes. The market value of a marginal vote at any time equals the expected
discounted equilibrium value of a vote at the time of a control contest. The
expectation is taken with respect to the probability of a differentiated bid-control
contest where a given marginal vote is actually sold to a contestant.
330
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
A simple example can be used to illustrate the valuation of a marginal vote.
Assume a widely held firm has a verifiable cash flow of one thousand dollars and one
hundred shares, half of which carry one vote each and the remaining half carry no
voting power. In the absence of private benefits, each share, regardless of its vote,
will trade at $10. Let the firm exist for two periods, no discounting. In period two,
there are two contestants for the control of that firm, who value control at $100, over
and above the verifiable cash flow of the firm. In this case the optimal controlwinning bid is $350 for 50% of the voting shares (25 shares with a total verifiable
cash flow of $250). An owner of a marginal voting share will receive $14 per share in
period two. Let the probability that such a contest happens in period two be 100%.
A voting share is demanded by a control contestant in period two with a 50%
probability. Therefore, in period one a vote is valued at $2, voting shares trade for
$12, whereas nonvoting shares are priced at $10.
Note that the value of a vote itself is not directly observable, but needs to be
backed out from the prices of multiple- and limited-voting shares. In this simple
example, the two classes of shares carry exactly the same verifiable cash flows per
share of $10. One could envisage that in reality each class pays a specific dividend,
different from that of the other class. The liquidity of the classes will differ as well,
thereby introducing further cash flow differences. Other potential controls are
discussed in Section 2.3. Once I control for such nonvote-related differences in the
characteristics of the share classes, I can use the market price of a nonvoting share as
a measure of the verifiable cash flows of a voting share, or $10. The premium value of
a voting share over and above $10 is interpreted as the expected discounted vote
value, $2. Possibly the most significant empirical challenge is to correctly control for
the probability of a differentiated-bid control contest. This correction permits us to
back out the per-share vote value at control contest time, $4. Aggregating vote
values across the control block is then trivial (25 votes are needed for control, valued
at $4 each), to arrive at the total value of control of $100.
In order to link the vote to control benefits as this example does, one needs to
assume a competitive control market, a firm with dispersed ownership, and a fixed
mechanism for control changes. In that case, the control contestants pay for control
up to their expected private benefits from control, and each vote demanded in the
control contest receives an equilibrium price equal to total private benefits per
control-block vote. Several vote-value studies adopt this interpretation. The
assumptions required for this result are restrictive, however, and the discussion at
the end of this section relaxes some of the assumptions A more conservative
viewpoint will claim that vote value permits the identification of a lower bound of the
benefits of control, since a control contestant would never pay for control more than
their expected control benefits.
In sum, the value of control-block votes, backed out from multiple- and limitedvoting share prices, is identified after controlling for the probability of the
votes being demanded during a control change, and nonvote-related sources of
share value, e.g., block-holding costs, dividend payment, and liquidity differences
between the shares classes. The vote-value measure and the controls are discussed
below.
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
331
2.1. The value of control-block votes
The sample includes all listed dual-class firms covered by DATASTREAM in the
30 countries with largest stock markets by 1995 market capitalization, 661 firms.
Dual-class firms are defined in Table 2. DATASTREAM is a financial information
service with wide international coverage of firms. Firm- level data is collected from
company sources (annual reports, filings with the national stock exchange, business
data gathering agencies).2
The value of control-block votes is calculated from the prices of multiple- and
limited-voting shares, adjusted to comprise 50% of the voting power, and scaled by
the firm market value. The computation formula is given in Table 2, and is derived in
Nenova (2001). There are three sample firms, all in Switzerland, with more than two
listed classes of shares, each class with different voting power. In those three cases, I
select the bearer class and the nonvoting class, and drop the registered class, in view
of ownership and transferability restrictions on the latter.
The vote value measure aggregates votes, instead of focusing on a single vote,
since comparisons of the value of a single vote are not informative across firms and
countries. In particular, the value of a marginal vote is not independent of the
concentration of voting power in each multiple-voting share, of the relative number
of shares in each class, and of the size of the corporation. To illustrate the first of the
three issues, take again the example of a firm where the votes of the control block are
worth $100. In the original example there are 50 voting and 50 nonvoting shares,
control is achieved with 25 votes, and each of those votes is valued at $4. Compare
that to a case where there are only ten voting and 90 nonvoting shares, so that
control is achieved with five votes and each vote is worth $20. Therefore, a correct
comparison requires the computation of the value of all votes in the control block.
This overall measure of the value of voting power needs to be further scaled, for
example, by firm market value, in order to get a quantity independent of firm size
and other characteristics. This approach focuses attention on the substantial
determinants of the value of control-block votes, rather than the mechanical
determinants such as the number of multiple-voting shares that the company
happens to have issued.
Table 3 summarizes the value of control-block votes as a share of total firm
value across all sample countries. Since some countries are overrepresented in
the sample, and can have undue influence on the results, I test using medians. Except
for the smallest country samples, the median value of control-block votes is
significantly different from zero in all countries. Median value of control-block
votes in Canada, Denmark, Finland, Sweden, and the US is below 1% of the total
value of the company. On the other end of the spectrum, median value of controlblock votes in Australia, France, Italy, South Korea, and Mexico is around
one quarter or more of company value. The value for Australia is an outlier for
common law countries; however, the estimate is based on three firms and should
be treated with caution. The simple average value of Hong Kong control-block
2
A list of all sources by country is available from the author upon request.
332
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Table 2
Description of variables
Dual-class firm
Total value of
control-block
votes as a share
of firm market
value
Rule of law
Investor
protection
Takeover
regulations
Charter
provisions
A dual-class firm complies with the following conditions. First, there are at least
two publicly traded share classes, with unequal voting per cash-flow rights.
Second, both classes are listed on a domestic exchange, and traded during the
period January 1 to December 31, 1997. Third, the limited-voting class is not
convertible into the multiple-voting class. Convertibility in the reverse direction
is allowed, since the multiple-voting class typically trades at a premium, so the
convertibility option from multiple-voting is out-of-the-money. Fourth, neither
class receives a fixed dividend, independent of that of the other class nor is
redeemable or callable at the option of the firm at a preset price (such securities
look like debt).
=[PM (t) PL (t)]/(1 k) * ([NM+NL*k] /2)/[NM*PM (t)+NL*PL (t)], averaged
over all t (52 weekly observations). PM (t) and PL (t) are the weekly prices of the
multiple and limited-voting shares, NM and NL are the number of multiple and
limited-voting shares, and k is the ratio of the voting power of a limited- to a
multiple-voting share. Prices are Friday close, January 1 to December 31, 1997,
from DATASTREAM, from the last trade or an official price fix from the main
domestic exchange. For shares not listed on the main exchange, I use regional
market prices. Prices are adjusted for subsequent capital changes. I
accommodate restructuring of security-voting arrangements during the
sample period by using only the data prior and up to the acquisition or share
withdrawal.
Assessment of the law and order tradition in the country produced by the
country risk-rating agency International Country Risk (ICR), 1982 to 1995.
Rescaled 0 to 1, with higher values indicating stricter rule of law. Source: La
Porta et al. (1998).
Averages six indicator variables of general investor protection: (1) 1 if
shareholders are allowed to mail their proxy votes, (2) 1 if the shares are
blocked before a general shareholder meeting, (3) 1 if cumulative voting is
allowed in the law, (4) 1 if oppressed minority provisions are incorporated in the
law, (5) 1 if shareholders enjoy a preemptive right over future issues, and (6) 1 if
the vote percentage required to convene an extraordinary shareholder meeting
is at or below 10%. Higher values indicating better protection. Source: La Porta
et al. (1998).
Averages three variables of investor protection during a corporate control
contest: (1) 1 if the legal code requires a control contestant to offer all classes
the same tender price, zero otherwise; (2) 1 if the legal code requires a buyer of a
large or majority block to pay minority shareholders the same price as for the
block shares by share class, zero otherwise; and (3) 1 minus the level of
ownership at which a dominant vote-owner is legally required to make an open
market bid for all shares, zero if there is no such provision in the law. Higher
values indicate stricter takeover law. Sources: International Society of Securities
Administrators (1997), Takeover code, Company Law, Securities Law,
Economist Intelligence Unit 1997.
Aggregates six charter provisions that concentrate power in the hands of
dominant shareholders: (1) golden shares, i.e., special decision-making rights of
dominant shareholders; (2) coattails, i.e., a provision that the limited-voting
shares become convertible into multiple-voting at the time of a control change;
(3) poison pills, i.e., provisions that make it costly for outsiders to purchase
large stakes; (4) voting caps, i.e., an upper limit on the votes that a single
shareholder can cast; (5) the right of the Board of Directors to limit transfer of
shares; and (6) the right of limited-voting shares to attain full voting power in
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
333
Table 2. (Continued)
Shapley value
Firm size
Dividend ratio
Cumulative LV
dividend
Registration
costs for MV
Log difference in
Turnover
case of dividend nonpayment, sale of assets, or excessive concentration of
voting power to cash-flow rights. Rescaled from 0 to 1, with higher values
indicating less power concentration with the controlling party. Source:
company filings, stock exchange publications, Moody’s.
The Shapley value measure is the continuous version for oceanic games (Milnor
and Shapley 1978). The Shapley value solution of a voting game reflects the
payoff or value of participating players, measured as the extent to which each
player is pivotal to the voted decision. The ownership data is for all
shareholders with 5% of the shares or more. Reporting requirements in some
countries are higher than 5% (e.g., 25% in Germany), however, the sources for
all sample countries usually identify all block owners. For example, in Germany
the median firm blockholders hold 75% of the voting power. Company
ownership structures data is from company filings, stock exchange publications,
Moody’s.
=ln {firm market value/stock market capitalization of the corresponding
country}.
The excess dividend payment to a limited- over a multiple-voting share, scaled
by the total dividend to a limited-voting share. In 37 cases firms do not
announce a dividend policy, so I extrapolate from past dividend history. Data
from Annual Reports, stock exchange data, Moody’s, Worldscope.
=1 if dividend payments to a limited-voting share are cumulative.
=1 if the multiple-voting class has registration costs which do not apply to the
limited-voting class.
=ln (limited-voting turnover/multiple-voting turnover). Turnover is the ratio
of the average daily trading volume of the class over all business days, to the
number of shares outstanding in 1996. The Datastream data refer to shares
traded on the main exchange, except in the US, where volumes are summed
over all exchanges on which the shares are listed. Sources: Datastream,
Bloomberg (Finland, South Africa).
votes is negative. The vote value measured in Table 3 is unadjusted and scaled
by the probability of a control change. All Hong Kong firms are majority owned,
and their probability of control change is low (or zero). Thus, the unadjusted
measure of vote value will be low (or zero). In fact, the Hong Kong vote value
is impossible to estimate after controlling for firm characteristics (this is done in
Table 4), since the probability measure used in this work, the Shapley value, has a
value of zero.
Table 3 presents the sample countries divided by legal origin, as defined in La
Porta et al. (1998). Median value of control-block votes is highest in French civil law
countries (22.6%), followed by German civil law countries (11.0%). Common law
and Scandinavian civil law countries score the lowest, with 1.6% and 0.5%,
respectively. Such low value of control-block votes in Scandinavian countries cans
seem surprising, considering the civil origin of the law. Agnblad et al. (2001) advance
the following explanation for the case of Sweden. They state that controlling parties
are law-abiding, so that ‘‘concerns about reputation and social status have a
334
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Table 3
Comparison of mean value of control-block votes to firm market value, raw figures
Mean and median value of control-block votes as a share of firm value, 661 firms, 1997, by legal origin.
Significance: a=1%; b=5%; c=10%.
Value of
control-block
votes/Firm
value
Mean
(two-tailed t-statistic)
Common law
Australia
Canada
Hong Kong
South Africa
UK
US
Common law average
French civil law
Brazil
Chile
France
Italy
Mexico
French origin average
Scandinavian civil law
Denmark
Finland
Norway
Sweden
Scandinavian origin average
German civil law
Germany
Korea (Rep. of)
Switzerland
0.2320c
(3.5384)
0.0276a
(3.2704)
0.0288
(1.9353)
0.0672b
(2.4586)
0.0957a
(5.0751)
0.0201b
(2.4051)
0.0447a
(6.3352)
0.2319a
(12.5585)
0.2314c
(1.6466)
0.2805a
(6.5915)
0.2936a
(13.8837)
0.3642a
(37.5729)
0.2539a
(18.2927)
0.0084c
(1.7556)
0.0503
(0.8749)
0.0583b
(2.9655)
0.0104b
(2.3211)
0.0048c
(0.4037)
0.0950a
(6.5319)
0.2894c
(1.9620)
0.0544b
(2.3675)
Median
Number
of
observations
0.2594
3
0.0047a
65
0.0152
5
0.0308a
22
0.0721a
27
0.0071a
39
0.0157a
161
0.1475a
141
0.0257
7
0.2747a
9
0.2993a
62
0.3652b
5
0.2261a
224
0.0029b
30
0.0052c
21
0.0438a
15
0.0043a
43
0.0052a
109
0.0493a
65
0.6000a
65
0.0147b
37
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
335
Table 3. (Continued)
0.1616a
(2.7780)
German origin average
0.1103a
167
Tests of equality of distributions (medians), T-statistics:
English vs. French
English vs. Scandinavian
English vs. German
11.732a
3.399a
6.571a
French vs. Scandinavian
French vs. German
German vs. Scandinavian
12.294a
1.751a
7.813a
disciplining effect on their behavior, discouraging minority abuse.’’ The value of
control-block votes in all four groups of countries is significantly different from zero.
Table 3 presents tests of equality of distributions among the value of control-block
votes in the four groups of countries. Median values of control-block votes in all
groups are significantly different from one another.
2.2. The probability of a control contest
The simple example at the beginning of this section illustrates the need to control
for the probability of the votes being demanded during a control change, as well as
nonvote-related sources of share value, in order to correctly back out the value of
control-block votes from multiple- and limited-voting share prices. This section
focuses on empirical difficulties in measuring and controlling for the probability of a
control contest. Section 2.3 discusses dividends, liquidity, and block-holding costs.
Section 2.4 addresses some further empirical problems such as survivorship bias and
the time behavior of vote value. Finally, Section 2.5 undertakes the estimation of the
value of control-block votes.
Traditionally, the vote-value literature proxies the probability of a takeover using
ownership-based variables. Studies of acquisition likelihood have advanced other
major determinants, however, in addition to ownership structure. These additional
determinants could include performance and firm- level characteristics (e.g., industry
indicators) and specifics of the takeover setup (e.g., takeover rules and takeover
defenses).
The probability of a control contest is inversely related to the ownership
concentration of the firm. A typical measure of the control-change probability
is the voting stake of the largest or several dominant vote-owners. That fails to
account for the number, composition, and interaction between the vote-holders. A
Herfindahl index controls for the number and composition of owners, but
the Shapley value measure in addition corrects for vote-holder interaction (Milnor
and Shapley, 1978). The reported results use the Shapley value control. I repeat
all tests with a Herfindahl index of ownership concentration, the voting stake
of the incumbent dominant vote-owner, the cumulative voting stake of the five
largest vote-owners, and a majority ownership dummy. The results are not sensitive
336
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Table 4
Cross-country estimation of mean value of control-block votes to firm market value
OLS regressions with country dummies, I–IV include all firms, V includes only firms with nonvoting
shares; VI includes only firms with equal dividends for both classes. The dependent variable is value of
control-block votes to firm market value. Variables are defined in Table 2. The controls are expressed in
deviations from their respective country means. Huber/White errors in parentheses. Significance: a=1%;
b=5%; c=10%. Data for 618 firms, 1997.
I.
Australia
Brazil
Canada
Chile
Denmark
Finland
France
Germany
Hong Kong
Italy
Korea (Rep. of)
Mexico
Norway
South Africa
Sweden
Switzerland
UK
US
Shapley value
Firm size
Log difference in turnover
Dividend ratio
0.2320a
(0.0356)
0.2301a
(0.0183)
0.0275a
(0.0081)
0.2314c
(0.1331)
0.0084
(0.0058)
0.0002
(0.0236)
0.2805a
(0.0399)
0.0949a
(0.0145)
0.0288b
(0.0132)
0.2887a
(0.0211)
0.4772a
(0.0669)
0.3642a
(0.0098)
0.0630a
(0.0200)
0.0672a
(0.0244)
0.0104b
(0.0046)
0.0561b
(0.0234)
0.0957a
(0.0172)
0.0201b
(0.0079)
0.0640b
(0.0307)
0.0024
(0.0041)
0.0152a
(0.0040)
0.2176a
(0.0548)
II.
0.2320a
(0.0373)
0.2301a
(0.0183)
0.0275a
(0.0080)
0.2314c
(0.1332)
0.0084
(0.0058)
0.0003
(0.0238)
0.2805a
(0.0406)
0.0949a
(0.0145)
0.0288b
(0.0132)
0.2888a
(0.0211)
0.4772a
(0.0670)
0.3642a
(0.0098)
0.0630a
(0.0201)
0.0672a
(0.0244)
0.0104b
(0.0046)
0.0561b
(0.0235)
0.0957a
(0.0172)
0.0201b
(0.0079)
0.0641b
(0.0307)
0.0023
(0.0041)
0.0153a
(0.0041)
0.2192a
(0.0552)
III.
0.2320a
(0.0355)
0.2301a
(0.0183)
0.0275a
(0.0081)
0.2314c
(0.1332)
0.0084
(0.0059)
0.0002
(0.0237)
0.2805a
(0.0401)
0.0949a
(0.0145)
0.0288b
(0.0132)
0.2887a
(0.0211)
0.4772a
(0.0670)
0.3642a
(0.0099)
0.0630a
(0.0201)
0.0672a
(0.0244)
0.0104b
(0.0046)
0.0567b
(0.0229)
0.0957a
(0.0172)
0.0201b
(0.0079)
0.0658b
(0.0309)
0.0024
(0.0041)
0.0152a
(0.0040)
0.2175a
(0.0549)
IV.
0.2320a
(0.0200)
0.2305a
(0.0179)
0.0323b
(0.0149)
0.1567
(0.1137)
0.2805a
(0.0395)
0.0949a
(0.0145)
0.2884a
(0.0212)
0.4758a
(0.0673)
0.3642a
(0.0104)
0.0625a
(0.0203)
0.0591a
(0.0177)
0.1215a
(0.0381)
0.1067a
(0.0210)
0.0162
(0.0125)
0.0884c
(0.0468)
0.0001
(0.0053)
0.0147a
(0.0041)
0.4986a
(0.0924)
V.
0.0251a
(0.0087)
0.2314c
(0.1336)
0.0092
(0.0068)
0.0207
(0.0205)
0.2665a
(0.0417)
0.0747a
(0.0252)
0.0550a
(0.0203)
0.0461a
(0.0096)
0.0594b
(0.0231)
0.0104b
(0.0051)
0.0568b
(0.0234)
0.0921a
(0.0179)
0.0164b
(0.0082)
0.0580b
(0.0251)
0.0096a
(0.0031)
0.3792
(0.4156)
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
337
Table 4. (Continued)
Cumulative dividend for limited-voting
0.0134
(0.0311)
Registration costs for multiple-voting
Number of Observations
Adjusted R squared
618
57.28%
618
57.21%
0.0326
(0.0370)
618
57.22%
425
59.37%
286
34.61%
to the choice of probability control. Alternatively, I use an extended set of
probability controls, which includes ROE, sales growth, leverage, and industry
dummies (not reported). This list of controls for probability is based on empirical
results from Palepu (1986), Billett (1996), Barber et al. (1999), Nuttall (1999), and
Cudd and Duggal (2000). While none of the extra measures are significant, the
results are maintained. The effect of takeover specifics on takeover probability is
captured in the tests below by the takeover variable. This effect is examined in
Malatesta and Walkling (1988), and Harris and Raviv (1988). As a second approach
to identifying the independent effect of takeover probability, I run a fixed effects
(within) regression to eliminate the effect of takeover rules (which are the same
within each country).
Walkling (1985) finds that a bidder’s toehold increases the probability of takeover
success. In this sample, the voting stake of the second largest stakeholder has an
insignificant effect on vote value. Meeker and Joy (1980) suggest that an unstable
controlling coalition will increase the likelihood of a change in control. I test this
proposition on coalitions among all family members who hold shares in the firm, but
the results lack significance (not reported). The tests are inconclusive, since the data
identifying coalition members in the sample countries is incomplete. The relation
between ownership and takeover likelihood could be nonlinear, since high vote
ownership also increases the control contest premium that the incumbent can afford
to pay (or bargain for). In this sample there are no significant nonlinearities, however
(tests not reported).
The Shapley value is defined in Table 2. A higher cumulative Shapley value of all
dispersed shares, i.e., a lower ownership concentration, increases the likelihood
of a control contest. The measure takes a value of zero in a majority controlled firm
and one in a widely held firm. In a firm with a single 30% owner the Shapley value
of the float is 0.57 instead of 0.70, capturing coordination costs of dispersed
voters. Another appealing property of the Shapley value of the free float is that
it increases as the stakes of the major shareholders become comparable. Shapley
value averages by country are given in the last column of Table 5 and range
from zero in Hong Kong (i.e., all companies are majority owned) to 0.73 in South
Korea, where there are legal restrictions on ownership. Shapley values are higher in
Anglo-Saxon and Scandinavian countries, since ownership is relatively more
dispersed.
Common law countries
Australia
20%
Canada
20%
Hong Kong
35%
South Africa
30%
UK
30%
US
None
Common law average
—
French tradition civil law countries
Brazil
None
Chile
None
France
33%
Italy
10%
Mexico
None
French average
—
Scandinavian tradition civil law countries
Denmark
50%
Finland
67%
Norway
45%
Sweden
None
Scandinavian average
—
German tradition civil law countries
Germany
50%
Korea (Rep. Of)
10%
Switzerland
33%
German average
—
Mandatory offer
threshold
0
1
1
1
1
0
0.83
0
0
1
0
0
0.20
1
0
1
1
0.75
0
0
0
0.00
0
0
0
0
0
0.00
1
1
0
0
0.50
0
0
0
0.00
In a tender offer, does
the minority get the
same price as the
control block, by class?
0
1
1
1
0
0
0.50
Same price for
all classes in a
tender offer?
0.13
0.23
0.17
0.17
0.63
0.33
0.39
0.25
0.40
0.00
0.00
0.42
0.23
0.00
0.13
0.20
0.70
0.66
0.68
0.43
0.00
0.44
Takeover
Rules Index
0.52
0.08
0.54
0.38
0.60
0.57
0.50
0.58
0.56
0.34
0.36
0.33
0.36
0.30
0.34
0.50
0.53
0.50
0.50
0.50
0.62
0.52
Charter
Provision Index
0.10
0.73
0.27
0.32
0.20
0.42
0.45
0.38
0.35
0.08
0.05
0.15
0.23
0.15
0.13
0.18
0.31
0.00
0.17
0.23
0.23
0.25
Shapley value
Table 5
Country-level legal environment and takeover provisions
The takeover rules index is defined in Table 2 and is based on the values of the indicators in columns 2 to 4. Charter rules and Shapley values are averaged over
all sample firms in the country (columns 6 and 7). The value of the charter provision index decreases with power concentration in the hands of the incumbent.
338
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
339
2.3. Block-holding costs and security-value differences between the share classes
A further determinant of the value of control-block votes are the costs associated
with financing a large block, participating in a control contest, and bearing
idiosyncratic risk when holding a large stake in the firm. Those costs increase with
firm size and are negatively related to the value of control-block votes. Firm size
could also have an independent effect on vote value since, for example, larger firms
are more scrutinized and it is costlier to extract private benefits. The measure of firm
size used here is company market value relative to the national stock market size
(defined in Table 2).
Differences in dividends and liquidity can cause mis-measurement of vote value.
For example, if the limited- voting class receives higher dividends than the multiplevoting class, the price difference between the two classes is smaller compared to the
case of equal dividend payments, all else constant. Definitions of the dividend and
liquidity controls are in Table 2. I also explore (and report some results) on foreign
ownership restrictions, tax treatment, costs of share registration, and convertibility
provisions for multiple- into limited-voting shares. Note that convertibility in the
opposite direction is not allowed in the sample.
2.4. Survivorship bias and time behavior of vote value
The sample represents a snapshot of dual-class companies at a point in time, which
raises concerns about survivorship bias. Specifically, companies that delist may
systematically differ in their vote value (e.g., the firms with most valuable control
eventually delist). To alleviate survivorship bias, the sample includes companies or
classes withdrawn from the market between January 1, 1997 and January 1999 (the
date of constructing the sample). If those firms systematically differ from the sample,
adding them back reduces the bias. A more compelling correction would be to use all
dual-class firms listed over a long period of time. DATASTREAM data availability
prior to 1995 is constrained to few countries, however.
This study adopts a cross-sectional perspective in its examination of the
relationship between the value of control-block votes and its legal and ownership
structure determinants, due to the time stability of both regulations and blockholdings. The lack of a time dimension begs the question of the time behavior of vote
value. The issue has not been extensively covered in the previous literature. However,
indicative results suggest that, on the country level, vote values are stable in time.
Megginson (1990) examines 152 listed dual-class UK firms during 1955 to 1982 and
finds that the vote premium is stable in time. Individual firm vote- values, on the
other hand, increase at the time of a contest over company control, due to the higher
probability of a marginal investor to realize the value of the votes owned. Rydqvist
(1996) presents data on the surge in vote value in four Swedish dual-class listed
companies that are subject to a takeover during 1983 to 1990. Proper control for the
probability of a takeover is called for. Country- level results will further alleviate the
problem, provided takeover activity is uncorrelated across firms and industries.
340
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
The second factor that is advanced in the literature as affecting the time behavior
of vote values are legal and regulatory changes. Smith and Amoako-Adu (1995) find
that around the time of the reform in Toronto Stock Exchange regulations, which
imposes mandatory coattail provisions for all listed firms in 1987, the average vote
value of listed Toronto firms increases from 4.2% in the period 1981 to 1986 to 6.4%
in the period 1998 to 1992. Nenova (2000) examines the jump in average control
value of listed Brazilian firms in response to the 1997 reform of the corporations law
and finds that control value doubles. Since this work focuses on the relation between
control value and laws, I assure the contemporaneous correspondence of the two
measures.
2.5. Estimation of the value of control-block votes
Table 3 does not control for takeover probability, block- holding costs, and
security value between the multiple- and limited- voting classes. Table 4 presents
adjusted vote value, and for comparability to Table 3, the controls are expressed as
deviations from the respective country means. The value of control-block votes is
significant in magnitude, and varies widely across countries. The highest average
value of control-block votes are in South Korea (48% of firm value) and Mexico
(36%). If one adopts the view that the value of control-block votes is at least as high
as the private benefits of control, then on average, South Korean controlling
shareholders appropriate close to one-half of the value of the company. Similarly,
high average value of control-block votes is found in Australia (23%), Brazil (23%),
Chile (16% to 23%), France (27% to 28%), and Italy (29%).3 In Scandinavian and
Anglo-Saxon countries, control-block votes are of very little value. In Denmark,
Finland, and Sweden, the value of control-block votes is 1% or less of the market
capitalization of firms. The value of control-block votes goes up to 4% in the US and
up to 7% in Canada. The Hong Kong estimate is discussed in detail above in the
comments on Table 3. The estimates are generally significant, with the exception of
Denmark and Finland, where vote value is almost zero.
The controls are of the expected signs, and significant, with the exception of firm
size. Alternative controls for security-value differences, namely the indicators for
cumulative dividend and for registration costs of multiple-voting shares, do not
prove to be of statistical significance, and render virtually unchanged results
(Regressions II and III). The results are similar if one includes indicators for
guaranteed minimum dividend for the limited-voting class, preference of limitedvoting shares, or convertibility for multiple-voting shares (not reported). The bid-ask
spread proxy for liquidity differences (not reported) limits severely the sample size
due to missing values, but otherwise renders similar results. The effect of foreign
ownership restrictions is insignificant in country-level tests (not reported). Countrywide foreign ownership restrictions on the multiple-voting class are in force in Brazil,
Chile, France, South Korea, and Mexico. Firms in Australia, Norway, and
Switzerland are allowed to impose foreign ownership restrictions at will. In the
3
Note that the estimate for Australia is based on 3 observations and therefore precision suffers.
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
341
regressions to follow, I keep only the excess dividend ratio and the liquidity controls
since they capture security value differences well.
The remaining regressions perform robustness checks. Regression IV restricts the
sample to firms whose limited-voting shares carry no vote. In this case, the
specification of the value of control-block votes measure hinges minimally on
assumptions. Regression V isolates firms with equal dividend policies for the two
classes. The coefficients remain stable.
Available vote- value measures in the literature are not strictly comparable to the
measure used in this work. As noted, existing studies measure the value of a single
vote, as opposed to the value of control-block votes. Thus Italy and the US have
vote-value measures of 82.7% and 12.5%, respectively, controlling for takeover
probability, dividends, liquidity, and firm size (Zingales 1994 and 1995).4 One needs
to make heavy assumptions in order to render these numbers comparable to the
results in Table 4. Going through such an exercise, the single vote-value measures
quoted above correspond to values of control-block votes of 39% for Italy and 2.9%
to 4.6% for the US. The corresponding measures in Table 4 are 28% and 1%.
The value of a single vote in the UK is 10.6% controlling for voting power relative
to cash-flow rights of the multiple-voting class, insider holdings, and firm size, but
not for liquidity (Megginson 1990).5 In terms of the value of control-block votes, this
could translate into anything from 2% of firm value (if the limited-voting shares
carry no voting rights) to 19% (if the voting power of multiple- to limited- voting
shares is 3:4, for example). Average vote ratios for the UK are unavailable. The
range of those estimates illustrates the demanding nature of the assumptions needed
in order to render these results comparable to the single-vote studies in the literature.
Using the value of a single vote instead of the value of all control-block votes impairs
comparability of results, as the above figures show. Studies of other countries do not
report sufficient information to compute comparable measures. Single vote values
are also available for Canada, 19.3% (median 6.4%), Korea, 9.6% (median 9.3%),
Sweden, 12%, and Switzerland, 1.2%.6
3. Determinants of the value of control-block votes
The results in Table 3 suggest a role for the legal environment in determining vote
value. This paper examines the effect of investor protection generally, takeover laws,
law enforcement, and corporate charter provisions that alter shareholder control
rights.
4
The US data is 1984 to 1990, and calculations use results in Regression III, Table 4 and summary
statistics in Table 2, assuming the ratio of voting power is in the range between one to five votes and zero
to one votes. For Italy, data is for 1987 to 1990, and calculations use Table 2, pooled regression, and the
summary statistics in Table 1.
5
Table 1, summary statistics and Table 4, regression with 1970 to 1977 prices.
6
The data for Canada is based on 1988 to 1992 data, Table 1, Smith and Amoako-Adu (1995); for
Korea, on 1992 to 1993 data, Table 3, Chung and Kim (1999); for Sweden, on 1983 to 1991 data, Table 1,
Rydqvist (1996); and for Switzerland, on 1983 data, Table 1, Horner (1988).
342
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
3.1. Law enforcement
Strict law enforcement and good general investor protection laws increase the
likelihood of a lawsuit and therefore a monetary or reputation loss to a controlling
party who is in a position to extract private benefits. That in turn decreases the
control premium that parties vying for control are willing to pay marginal voteowners, implying a negative relationship between better laws and the value of
control-block votes. I use the measure for rule of law from La Porta et al. (1998), and
define it in Table 2. The results are stable when using alternative measures of law
enforcement suggested by La Porta et al. (1998) which include efficiency of the
judiciary, accounting standards, and corruption.
3.2. Investor protection
Investor protection provisions, such as disclosure and accounting standards, or
the availability of an oppressed minority mechanism, uphold minority investor value
and are thus negatively related to the value of the control-block votes. However,
some sections of the law can be expressly directed at one class (e.g., dividend pay-out
rules or voting and representation rules). The effects of such rules on the voting
premium can be ambiguous, since there are two forces at work. First, investor
protection curbs value transfers from minority shares of that class to the controlling
entity, which translates into a lower vote value. Second, there may be a certain
degree of substitutability of value transfers between the two share classes, in the
sense that the less protected class shoulders some of the value loss experienced by the
controlling entity as a result of the tighter protection of the other class. This
argument has not been explored in the literature; however, if this phenomenon is
observed in practice, there is a strong case for tightening minority protection for all
classes simultaneously. In the special case of both high substitutability and lower
protection of the limited-voting class, vote value can increase with investor
protection. I use the measure for anti-director rights from La Porta et al. (1998)
and define it in Table 2.
3.3. Takeover law
Takeover regulations ensure that the control premium is shared with minorities in
the event of control transfer. I concentrate on three characteristics of takeover
regulation: (1) equal treatment of share classes in a takeover; (2) equal treatment of
minority shares and the dominant block, within each class, in a takeover; and (3)
mandatory offers triggered above a certain ownership threshold (Table 5).
A large proportion of countries that allow dual-class shares have adopted equal
treatment of share classes in control changes. Equal price provisions between classes
can cause a zero vote value to be observed, even if the value of control-block votes is
strictly positive. In practice, this is not the case, because the equal price rule does not
apply for all forms of control transfer, thus leaving a positive probability for a
differentiated offer at control contest time (Smith and Amoako-Adu, 1995). On the
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
343
other hand, fear of lawsuits can decrease the observed instances of differentiated bids
in countries where those are permitted. Differentiated bids are allowed both in the
UK and US. In the US only four out of 30 takeover bids during the period of 1960 to
1980 offer unequal prices to each class (DeAngelo and DeAngelo, 1985). In the UK,
differentiated bids are frequent. In the period 1955 to 1982, 37 out the 43 takeovers
involved a differentiated offer (Megginson, 1990).
The fair price provision prescribes an equal price for minority shares and the
dominant block, within each class, in the case of a control transfer. The provision is
in force in four out of six common law countries and four out of 12 civil law (mostly
Scandinavian) countries. Fair pricing need not be beneficial. If takeover activity is
sufficiently valuable to minority multiple- or limited- voting investors, then stricter
enforcement of fair-pricing rules that raises takeover costs can hurt investors.7
However, markets for control are relatively inactive in this sample, casting doubt on
the empirical importance of this effect. Note that in the absence of a protective law,
fair pricing to minority shareholders can still occur under the threat of litigation, as
the following example shows. In June 1996, Mahle GmbH bought 51% of the voting
power in Metal Leve SA, Brazil, from the owner-families in a private deal. In
September the same year, Mahle GmbH offered to purchase the free float voting
shares (16.8% of the votes) on the open market, at the same price. Fair price offers
are infrequent in Brazil, since the control transaction terms are not usually disclosed.
In the case of Mahle GmbH and Metal Leve SA however, both companies had
subsidiaries in the US with a large share of the piston and engine part market. The
US Federal Trade Commission publicized the terms of the control transfer while
investigating the matter, possibly affecting the decision of Mahle GmbH to treat the
minority multiple- voting shareholders in a fair manner.
In all but five sample countries the takeover code imposes on the dominant voteholder to make a mandatory offer for all shares once a certain vote level is attained.
For example, in South Africa the Securities Regulation Code on Takeovers and
Mergers 1991 sets that threshold at 30% of the votes.
Takeover regulations increase the cost of control by compelling the new
controlling party to share more value at takeover time with minority shareholders
(of either class depending on each of the three regulations, see above). As with
enforcement and investor protection laws, I would except a negative relationship
between the strictness of takeover rules and the value of control-block votes. There
are three potential caveats for the enumerated takeover rules. First, the principle of
protecting shareholders in a takeover can be maintained by a judge even in the
absence of an explicit law, as illustrated in the examples with equal pricing. Second,
enforcement of the law can differ. For example, South Africa’s Rules of the
Securities Regulatory Panel feature rule 34, that permits the panel to exempt
investors from any control requirement that is transfer-related. Third, takeover rules
change the very mechanism of the control contest, in addition to increasing the cost
of a control change. A more careful approach would model faithfully the legal
specifics of the environment in the spirit of Bebchuk (1994).
7
I thank a referee for raising this point.
344
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
The takeover rules index is defined in Table 2. Table 5 shows that it is highest for
the common law countries (with an average of 0.45), followed by the Scandinavian
civil law countries (0.40), the German civil law countries (0.17), and the French civil
law countries with the lowest average (0.13).
3.4. Corporate charter provisions
Corporate charter provisions can be used to concentrate control in the hands of
dominant shareholders, over and above control afforded by voting power. I collect
information on golden shares, voting caps, limitations on transfer of shares, and
poison pills. For example Stora Kopparberg Bergslags AB, Sweden, is owned 18%
and controlled 22% by Mr. Wallenberg. The voting power of every other
shareholder is capped at 10%, thus giving the dominant owner the upper hand.
Due to the presence of a voting cap, the value of the charter provision variable for
Stora will be lower, signifying the presence of power-concentration by the
controlling owner. Data are gathered also on two provisions that abate the
inequality of power between the dominant party and other shareholders, namely
coattail provisions and provisions for enhanced voting rights of limited-voting shares
in case of dividend nonpayment, sale of assets, or excessive concentration of voting
power in the multiple-voting class. Control-concentrating charter provisions make it
cheaper to maintain corporate control. The value of control-block votes is expected
to increase with control-concentrating charter provisions and decrease with
provisions that protect limited-voting shares. Charter provision adoption can
intensify as a response to a strict takeover code. Further, charter provision as well as
takeover codes could affect the likelihood of a control challenge (Harris and Raviv,
1988). These considerations of endogeneity are addressed below.
The charter rules index is defined in Table 2. Country means for the index are in
Table 5. The index is lowest in the French and German civil law group of countries,
with an average of 0.34 and 0.38, respectively. The index is the highest in the
common law countries (0.52) and the Scandinavian countries (0.56). Civil law
countries tend to use of the voting caps and limits on transfers of shares. Common
law countries tend to use more extensively poison pills, golden shares, coattails, and
increased voting power in certain events.
To sum up, the value of control-block votes is expected to decrease with the
strictness of the legal environment. In particular such strictness includes better
general investor protection, higher quality of law enforcement, and stricter takeover
laws. The value of the control-block votes increases with the existence of charter
provisions that concentrate power in the dominant vote-owner.
3.5. Results
Table 6 explores the link between laws and the value of control-block votes.
Regressions I–IV show a negative and significant effect of tougher rule of law,
stricter general investor protection laws, investor- friendly takeover laws, and the
absence of power-concentrating charter provisions on the value of the control-block
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
345
votes. The magnitude of the effect is important as well. For example, going from the
most lax to the toughest takeover- law score reduces the value of control-block votes
of a widely held company from 23% to 10%. A move from full to median use of
power-concentrating charter provisions shaves 17 percentage points off the value of
control-block votes of a widely held company.
Regressions V and VI combine the four legal measures. The coefficients of the
legal variables generally remain significant. The magnitudes of the effects are
somewhat reduced, but remain considerable, with the exception of the takeover rules
variable. Note that the legal measures are significantly correlated. The highest
correlations are between enforcement and takeover rules (31.38), takeover and
general investor protections (38.18%), and charter provisions and enforcement
(47.82%). The four legal variables are strongly jointly significant, with an F-statistic
of 21.04. A potential econometric problem with Regression V is omitted variable
bias. Ideally, a random effects regression would allay our worries about an omitted
country-specific variable. The Hausman test suggests that the random effects
estimator is invalid, however. Therefore, Panel A presents OLS firm- level
regressions with clustered robust standard errors, to correct for possible withincountry correlation. Clustered robust standard errors are similar to the Huber-White
errors, but the weights are sums over each cluster, in this case each country. This
adjustment corrects the standard errors, but fails to correct the coefficients. As an
alternative test, Regression VI includes country dummies to gauge omitted countrylevel effects, and the results confirm those of Regression V. Similar conclusions can
be made when repeating Regression VI with indicator controls for legal origin, or
GNP per capita (not reported).
The effect of the legal measures is complementary, but proper enforcement is more
powerful than both investor protection laws and takeover regulation. From
Regression VI, the average widely held firm in a country with a ‘‘worst case’’ legal
protection enjoys a value of control-block votes of up to 48%. This figure assumes
no dividend or liquidity differences among the classes of shares, and uses the lowest
in-sample value for the legal environment variables. Raising the value of each legal
indicator in turn to the maximum in-sample value is illustrative of the link between
control value and laws. The value of control-block votes drops to 31% if law
enforcement quality is radically overhauled (from a minimum of 0.44 to a maximum
of one). A similar improvement in general investor protection (from 0.16 to 0.83)
drives the value of control-block votes further down to 20%. This figure falls to 8%
after tightening takeover laws (from zero to 0.7), and finally to 5% when I move
from full use of power-concentrating charter provisions to none (0.08 to 0.62). This
simple exercise reveals the considerable magnitude of the effect of laws, and their
proper enforcement, on the relative value of minority multiple- and limited-voting
shares.
Regressions I–VI of Table 6 assume that charter rules and takeover probability are
exogenous. In reality, it can be argued that they both depend on the takeover rules
that are in force in the country. In addition, charter provisions can affect takeover
probability. The correlation of takeover probability with takeover laws and with
charter rules are 15.80% and 9.29% and are not significant, however. Inasmuch as
Dividend ratio
Log difference in turnover
Firm size
Shapley value
Charter provisions
Takeover regulations
Investor protection
Rule of law
Panel A
b
0.1024b
(0.0474)
0.0031
(0.0039)
0.0154a
(0.0036)
0.2603c
(0.1377)
0.4524
(0.0903)
I.
0.0998
(0.0670)
0.0009
(0.0048)
0.0112a
(0.0022)
0.4564b
(0.1789)
0.2325b
(0.0648)
II.
0.1145
(0.0771)
0.0045
(0.0068)
0.0139a
(0.0029)
0.4094b
(0.1691)
0.1870b
(0. 0653)
III.
0.3469b
(0.1213)
0.0664
(0.0409)
0.0039
(0.0036)
0.3245b
(0.1336)
0.0136a
(0.0015)
IV.
a
0.2978
(0.0942)
0.1677b
(0.0709)
0.0330
(0.0558)
0.2325b
(0.1077)
0.0900a
(0.0274)
0.0017
(0.0035)
0.2001c
(0.1066)
0.0137a
(0.0037)
V.
0.1781
(0.1303)
0.3454c
(0.1830)
0.2616c
(0.1404)
0.1665b
(0.0804)
0.0562c
(0.0290)
0.0017
(0.0042)
0.2253a
(0.0561)
0.0148a
(0.0038)
VI.
0.1628a
(0.0329)
0.0603b
(0.0244)
0.0016
(0.0034)
0.2098a
(0.0594)
0.0149
(0.0127)
VII.
Table 6
Institutional determinants of mean value of control-block votes to firm market value
Panel A presents OLS regressions with clustered (robust) standard errors (I–VI). Regression VII presents fixed-effects (within) estimators. Regression VI is with
country dummies, which are not reported. Panel B presents (between) regressions of means value of control-block votes to firm market value, by country, on
the legal variables. Rule of law, investor protection and takeover regulations are scaled 0-1 with higher values for better laws. Charter provisions are scaled 0-1
and lower values denote power concentration with the incumbent. Shapley values are scaled 0-1. Firm market value is relative to stock market capitalization.
The dividend ratio is excess dividend of limited- over multiple-voting shares to total limited-voting dividend. Turnover is the log of limited- to multiple-voting
average daily trading volume in 1996. Standard errors in parentheses. Significance: a=1%; b=5%; c=10%. Data are for 1997.
346
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Number of Observations
Adjusted R squared
Constant
Charter provisions (country average)
Takeover regulations
Investor protection
Rule of law
Panel B—Between Regressions
Number of observations
Adjusted R squared
Constant
b
0.2762a
(0.1226)
18
25.69%
0.3724
(0.1420)
I.
0.4618a
(0.0820)
618
31.68%
0.2405a
(0.0699)
18
11.40%
0. 2082c
(0.1166)
II.
0.2200b
(0.0699)
618
30.52%
0.2076a
(0.0441)
18
20.68%
0.2688b
(0.1153)
III.
0.1130c
(0.0639)
618
16.91%
b
0.5326a
(0.1143)
18
44.94%
0.3318
(0.1232)
0.1273
(0.0987)
0.1879c
(0.1037)
IV.
0.2405a
(0.0550)
618
27.62%
0.0119
(0.1330)
0.0434
(0.0787)
0.1341
(0.0801)
0.6770a
(0.2005)
0.5102a
(0.0868)
18
68.42%
V.
0.5675a
(0.0857)
618
35.62%
0.6944a
(0.0844)
618
41.42%
0.2034a
(0.0288)
618
6.61%
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
347
348
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
the charter rules and the takeover probability are determined by prevailing laws,
their significance in the regressions could result from a cross-country correlation with
the value of control-block votes, which is caused by the dependence of all three
variables on the legal framework. I run a fixed-effects (within) regression, to
determine a lower bound for the explanatory power of takeover probability and
charter provisions for the value of control-block votes (VII, Table 6). Within
countries, the legal framework is held constant, so takeover probability and charter
provisions are exogenous factors in this setup. Both takeover probability and charter
provisions are significant within-country determinants of the value of control-block
votes.
A potential problem with the results in Panel A of Table 6 is that countries with
many observations have a disproportionate weight in the regression results. Thus, it
is difficult to judge whether the significance of the results comes from the fact that
they hold for the countries with most observations, or because the relations generally
hold. I therefore employ two additional sets of regressions as a robustness check. I
run country by country regressions for all countries with more than 10 observations
with similar results (not reported). As a further robustness check, Panel B of Table 6
shows (between) regressions, or OLS of mean value of control-block votes, by
country, on the four country-level legal measures. The four legal environment
variables together explain 68% of the cross-country variation in the value of controlblock votes. Note that the R squared is not affected by the collinearity problem. This
large effect of laws has a very attractive interpretation if one is willing to accept the
empirical validity of the relationship between vote and control value. In that light,
the results show the importance of the legal framework in curbing private benefits.
Adopting the more conservative empirical view, the results show that minority
shareholder rights are very effectively upheld with the help of a refined legal
environment, especially the rights of limited-voting shareholders which are
inherently less protected due to their limited (or nonexistent) voting power.
4. Conclusion
The value of control-block votes is an important part of firm value and varies
widely across countries. Average value of control-block votes ranges from 48% of
firm value in South Korea to 2.88% (assuming the controls are at their mean
values) in Hong Kong. Quite a few countries exhibit high vote values. The value of
control-block votes in Brazil, Chile, France, Italy, and Mexico are one-quarter or
more of firm market capitalization. These magnitudes are even more striking if one
considers that control-block votes value is a lower bound for the value of control.
Scandinavian and Anglo-Saxon countries, on the other hand, reveal a value of
control-block votes of less than 10%.
A value of control-block votes ranging up to one-half of firm market
capitalization brings to the fore the issue of a corporate governance mechanism
capable of protecting minority shareholder value, especially that of limited-voting
shares. Law enforcement, investor protection, takeover regulations, and corporate
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
349
charter provisions together explain 68% of the systematic differences in the value of
control-block votes. This large effect of laws has a very attractive interpretation if
one is willing to accept the empirical validity of the relationship between vote and
control value. In that light, the results show the importance of the legal framework in
curbing private benefits, underscoring accompanying implications about stock
market development, external financing, and corporate valuation. Adopting the
more conservative empirical view, the results show that minority shareholder rights
are very effectively upheld by a refined legal framework.
Vote value studies provide important insight into minority protection issues and
enable cross-country comparisons. The paper provides the first consistent crosscountry study of vote value by transferring the focus away from the value of a single
vote towards the control block as a measure of the value of voting power. The legal
framework that governs the link between those two magnitudes is made explicit. The
paper provides a comprehensive description of relevant corporate control market
regulations, as well as a taxonomy of firm- level corporate charter provisions.
References
Agnblad, J., Berglof, E., Hogfeldt, P., Svancar, H., 2001. Ownership and control in Sweden: Strong
owners, weak minorities, and social control. In: Barca, F., Becht, M. (Eds.), The Control of Corporate
Europe. Oxford University Press, London, pp. 228–258.
Barber, B., Palmer, D., Wallace, J., 1999. Determinants of conglomerate and predatory acquisitions:
Evidence from the 1960s. Journal of Corporate Finance 1, 283–318.
Barclay, M., Holderness, C., 1989. Private benefits from control of public corporations. Journal of
Financial Economics 25, 371–395.
Bebchuk, L., 1994. Efficient and inefficient sides of corporate control. Quarterly Journal of Economics
109, 957–993.
Bebchuk, L., 1999. A rent-protection theory of corporate ownership and control. NBER working paper
7203. National Bureau of Economics Research, Cambridge, MA.
Bergstrom, C., Rydqvist, K., 1990. Ownership of equity in dual-class firms. Journal of Banking and
Finance 14, 255–269.
Bergstrom, C., Rydqvist, K., 1992. Differentiated bids for voting and restricted voting shares in public
tender offers. Journal of Banking and Finance 16, 97–114.
Berle, A., Means, G., 1932. The modern corporation and private property. Macmillan, New York.
Billett, M., 1996. Targeting capital structure: the relationship between risky debt and the firm’s likelihood
of being acquired. Journal of Business 69, 173–192.
Chung, K., Kim, J., 1999. Corporate ownership and the value of a vote in an emerging market. Journal of
Corporate Finance 5, 35–54.
Claessens, C., Djankov, S., Lang, L., 2000. The separation of ownership and control in east asian
corporations. Journal of Financial Economics 58, 81–112.
Cudd, M., Duggal, R., 2000. Industry distributional characteristics of financial ratios: an acquisition
theory application. Financial Review 35, 105–120.
DeAngelo, H., DeAngelo, L., 1985. Managerial ownership of voting rights. Journal of Financial
Economics 14, 33–69.
Dodd, P., Warner, J., 1983. On corporate governance: a study of proxy contests. Journal of Financial
Economics 11, 401–439.
Dyck, A., Zingales, L., 2002. Private benefits of control: An international comparison. NBER working
paper 8711. National Bureau of Economics Research, Cambridge, MA.
350
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
Grossman, S., Hart, O., 1988. One share-one vote and the market for corporate control. Journal of
Financial Economics 20, 175–202.
Harris, M., Raviv, A., 1988. Corporate control contests and capital structure. Journal of Financial
Economics 20, 55–86.
Horner, M., 1988. The value of the corporate voting right: Evidence from Switzerland. Journal of Banking
and Finance 12, 69–83.
International Society of Securities Administrators, 1997. ISSA Handbook. Zurich, Switzerland.
Jensen, M., Meckling, W., 1976. Theory of the firm: Managerial behavior, agency costs, and ownership
structure. Journal of Financial Economics 3, 305–360.
Jog, V., Riding, A., 1986. Price effects of dual-class shares. Financial Analysis Journal 42, 58–67.
Johnson, S., La Porta, R., Lopez-de-Silanes, F., Shleifer, 2000. Tunnelling. American Economic Review
90, 22–27.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1997. Legal determinants of external finance.
Journal of Finance 52, 1131–1150.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and finance. Journal of Political
Economy 106, 1113–1155.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000a. Agency problems and dividend policies
around the world. Journal of Finance 55, 1–33.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000b. Investor protection and corporate
governance. Journal of Financial Economics 58, 3–27.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2002. Investor protection and corporate
valuation. Journal of Finance 57, 1147–1170.
Lease, R., McConnell, J., Mikkelson, W., 1983. The market value of control on publicly traded
corporations. Journal of Financial Economics 11, 439–471.
Levy, H., 1982. Economic valuation of voting power of common stock. Journal of Finance 38, 79–93.
Malatesta, P., Walkling, R., 1988. Poison pill securities: Stockholder wealth, profitability, and ownership
structure. Journal of Financial Economics 20, 347–376.
Meeker, L., Joy, M., 1980. Price premiums for controlling shares of closely held bank stock. Journal of
Business 53, 297–314.
Megginson, W., 1990. Restricted voting stock, acquisition premiums, and the market value of corporate
control. The Financial Review 25, 175–198.
Milnor, J., Shapley, L., 1978. Values of large games II: Oceanic games. Mathematics of Operations
Research 3, 290–307.
Modigliani, F., Perotti, E., 1998. Protection of minority interest and the development of security markets.
Managerial and Decision Economics 18, 519–528.
Nenova, T., 2000. Changes in corporate law in Brazil and the value of control. Unpublished working
paper. Harvard University, Cambridge, MA.
Nenova, T., 2001. How to dominate a firm with valuable control: Regulation, security-voting structure,
and ownership patterns of dual-class firms. Unpublished working paper. Harvard University,
Cambridge, MA.
Nuttall, R., 1999. Takeover likelihood models for UK Quoted Companies. Unpublished working paper
No. 6. Nuffield College, University of Oxford.
Palepu, K., 1986. Predicting takeover targets: A methodological and empirical analysis. Journal of
Accounting and Economics 8, 3–35.
Robinson, C., White A., 1995.The value of a vote in the market for corporate control. Unpublished
working paper, York University, Toronto, Canada.
Rydqvist, K., 1996. Takeover bids and the relative prices of shares that differ in their voting rights. Journal
of Banking and Finance 20, 1407–1425.
Shleifer, A., Vishny, R., 1997. A survey of corporate governance. Journal of Finance 52, 737–783.
Shleifer, A., Wolfenzon, D., 2002. Investor protection and equity markets. Journal of Financial
Economics 66, 3–28.
Smith, B., Amoako-Adu, B., 1995. Relative prices of dual class shares. Journal of Financial and
Quantitative Analysis 30, 223–239.
T. Nenova / Journal of Financial Economics 68 (2003) 325–351
351
Walkling, R., 1985. Predicting tender offer success: a logistic analysis. Journal of Financial and
Quantitative Analysis 20, 481–478.
Zingales, L., 1994. The value of the voting right: a study of the milan stock exchange experience. The
Review of Financial Studies 7, 125–148.
Zingales, L., 1995. What determines the value of corporate votes. The Quarterly Journal of Economics
110, 1047–1073.